Perhaps the most elegant demonstration of this occurred when a software salesman was demonstrating "Structural optimization software". Structural optimization software is a computer program where the modeler builds a very rudimentary "tinker toy" model of the structure and the software adds material where it is highly stressed and whittles it away where it not needed. By coding this process into the software, the computer can run hundreds, perhaps thousands of trials to find the "optimum" structure.
The typical patter for one of these sales dog-and-pony shows is to present a problem that "everybody" knows the answer to, seed the process with a highly non-optimal solution and then verify the software's excellence by showing the (expected) end results.
The problem this particular sales person chose was a diving board. The load was applied up-down and side-to-side. The seed was a constant round section. Everybody knew that the software would deliver a round section that tapered with the fat end bricked into the side of the swimming pool.
And everybody was wrong. Not about the tapered part. They (we) were wrong about the round part.
The section morphed from round to "iron cross". The salesman rather apologetically observed that optimization software was very unforgiving if the user forgot to input constraints or secondary loads (like torsion). In fact, this was an excellent salesperson. He pointed out that software is unhampered by biases and preformed conclusions. It found a solution that nobody had considered.
I remember thinking at the time that had computers not been invented, one could have optimized structures (like engine connecting rods) by keeping thigh bones alive in a culture (or maybe even within the pig!) and repeatedly applying the appropriate loads to them. Bones "build" where stressed and slough where not stressed. Forensic anthropologists can measure bones and determine if the deceased dragged stone to build pyramids, pulled oars on slave galleys or chased gazelles across the steppes. Bones tell no lies.
I am reminded of the structural optimization software story whenever I see a tree with "winged" twigs.
What does a tree require, as an organism, to survive a fire? It requires a few viable buds (but not all of them) somewhere on each twig. It requires at least one continuous path of undamaged cambium (inner bark) back to the roots to transport carbohydrates. While it would seem obvious to make the bark thicker everywhere, evolution (or God) determined that it was more economical to budget a heavier coating of insulation over a smaller portion of the circumference of the twig. The organism does not require 360 degrees of the circumference the entire length of the twig...it only needs three or four cells width as long as it is continuous from the buds to the roots.
Furthermore, the organism strategically invests resources in protecting those places most likely to harbor a bud. It packs a little bit of extra insulation at the "joint" between first and second year wood, and the joint where second and third year wood meet. That is where the latent buds reside.
The protuberances of "cork" on a Burr Oak twig appear highly random until you start thinking about it from the organism's perspective. Then it seems coldly, clinically and strategically.....perfect.
The Burr Oaks and Rock Elms and Euonymus of the world exhibit genius at the organism level. Not only do they buy life insurance by parsimoniously packing insulation at the very most important places, they practice triage and sacrifice the many to save the critical few.
Every bud does not need to live. Every cell of the cambium needs to live. Better to sacrifice the 80% if it guarantees that the critical 10% lives.
It goes beyond that. They are geniuses at the species level. These species produce swarms of progeny that exhibit wildly varying degrees of these characteristics.
A gambler can "win" at every horse race if he bets on every horse. The key to winning money (as opposed to merely "winning") is to bet more on the most likely winners and lesser amounts on the tails of the bell-shaped curve.
It goes beyond twig shape
Consider seed size: Acorns of remnant trees from Oak Savanna tend to be quite small for the species. The most likely "vector" for spreading these acorns are Blue Jays and Passenger Pigeons. Thick grass, long distances between trees and heavy carnivore populations make squirrels a poor agents of dispersal. We cannot know the preferences of Passenger Pigeons, but we know that Blue Jays are highly biased against acorns larger than 15mm in diameter. It is just too difficult for them to carry larger acorns. The acorn is much too likely to slip from their bill (like a watermelon seed when squeezed) before they get back to their preferred perch for eating.
Acorns from savanna remnant Burr Oaks in Michigan tend to range between 15mm and 20mm. Still small enough to tempt birds but large enough that some of them will slip out of the bird's bill and pioneer new ground.
Studies of Burr Oak growing in bottom lands show that the acorns are much larger. These trees were most likely spread by mammals, including humans. Suppose you were packing supplies for a long canoe trip. Would you pick 2.4 ounce Snickers bars or itty-bitty, individually wrapped, party favor 0.5 ounce candies?
Same species. Same zipcode. Factor of three or four in acorn size. Nature, God, continues to amaze.